The Signal
HRL’s SpinQICK Runs a Fully Connected Spin Qubit
A paper by HRL Laboratories researchers reports coherent control, state preparation and readout of a fully connected triangular spin-qubit device using SpinQICK. The authors also use dynamical-decoupling measurements to examine noise and dr
Quantum.world Team · Written with AI · fact-checked ·
98.9% of measured counts fell in the singlet category
HRL Laboratories researchers used an open-source radio-frequency system-on-chip control stack to prepare and read out a fully connected semiconductor spin qubit. In a paper submitted to arXiv on October 22, 2026, they report that 98.9% of counts after singlet preparation were categorized as singlets at the readout window. The number describes a specific state-preparation-and-measurement result, not a general gate-fidelity score.
The control package, SpinQICK, runs on an AMD-Xilinx ZCU216 RFSoC evaluation board and builds on the Quantum Instrumentation Control Kit, or QICK. The researchers used custom firmware and commercially available amplifiers to control a triangular triple-quantum-dot device, where three electrons interact through exchange coupling between each pair. The SpinQICK code is publicly available.
Three exchange axes test the control stack
Each exchange interaction lets the researchers rotate the encoded qubit about a different axis. Unlike a linear arrangement of three dots, the triangle adds a third coupling, allowing the team to test control across all three links and to apply a dynamical-decoupling sequence they call NZM1. In the paper’s experiment, modified firmware included crosstalk compensation; the SpinQICK repository README lists DSP crosstalk compensation as a near-term update to the package.
The ZCU216 supplied high-speed voltage pulses and digitized the charge-sensor signal for readout. The paper describes 14 baseband pulse channels and two channels configured for readout-signal generation; its control program schedules pulses and readout on a 2.5-nanosecond clock. The board’s digital signal processing demodulates and integrates the readout signal on the device, reducing the data that must be transferred for later analysis.
Decoupling experiments expose noise and drift
The team used two decoupling sequences to distinguish noise sources. Their NZM1 sequence, enabled by the triangle’s third exchange axis, suppresses quasistatic hyperfine noise while retaining sensitivity to exchange noise. Comparing the measurements with simulations helped the authors assess whether the control timing and crosstalk compensation supported the experiment.
One measured coherence lifetime was about 60 microseconds, roughly one-tenth of the approximately 600-microsecond result cited from earlier work on a device with similar intrinsic noise. The authors attribute the difference to a fitted high-frequency white-noise floor of about 40 nanovolts per square root hertz. They say the room-temperature amplifier chain likely contributes significantly, but the paper does not establish it as the sole cause.
Separate NZM1 measurements showed return-fidelity oscillations that the simulations did not reproduce. The researchers associate those oscillations with slow, correlated fluctuations and report calibration drift during data collection. Because the experiment used the same calibrated pulse voltages throughout that collection, the result highlights a practical limit alongside the control demonstration: a system can execute calibrated pulses while its operating point shifts over time.
Open software, with engineering work left to do
The paper presents SpinQICK as a lower-cost, adaptable alternative to proprietary control equipment, but it does not show that the system matches the stability or performance of mature commercial controllers. The authors also identify the amplifier chain as an area for improvement, particularly given the white-noise floor and the shorter measured lifetime.
“Although long-timescale stability remains an avenue for improvement, the results presented here establish spinQICK as an accessible and effective control solution for semiconductor spin qubits.”
Abigail L. Wessels and co-authors, researchers who authored the paper, write.
The experiment demonstrates coherent control, state preparation and readout on one fully connected triangular device; it does not report a two-qubit gate demonstration or benchmark a processor. The paper’s concrete engineering test for the next iteration is whether a lower-noise amplification chain can improve coherence without sacrificing the fast exchange pulses used in the experiment.